Concept and feasibility of the Augsburg Longitudinal Plasma Study (ALPS) – a prospective trial for comprehensive liquid biopsy-based longitudinal monitoring of solid cancer patients

Study protocol and patient recruitment

Patient enrollment and prospective biobanking is performed after patients have provided written informed consent for the ALPS trial and simultaneously for biobanking at the Augsburg Central Biobank (ACBB). The essential feature of the study is that it accompanies the patient in routine clinical practice, that all additional examinations are performed as part of the clinical routine examinations, and that ALPS thus does not alter the clinical course of the patient. Main inclusion and exclusion criteria are given in Supplementary Methods. Eligible patients are identified through reviewing appointment lists, recommendations from tumor conferences and direct contact with the treating physicians. This trial is conducted in accordance with the Declaration of Helsinki, approved by the Local Ethics Committee and registered at clinicaltrials.gov (ClinicalTrials.gov Identifier: NCT05245136).

Sample collection, preparation, and storage

Peripheral blood is obtained by collecting five and three 9 mL EDTA containers at baseline and at each follow-up visit in addition to regular blood tests, respectively. Blood samples are processed within 2 h of collection. Plasma is separated by centrifugation at 2,000 g for 10 min (Hettich Rotina 38R, 20 °C, no brake), transferred to a new tube, and centrifuged at 2,000 g for 10 min at room temperature to eliminate any residual cellular debris. Plasma samples are then stored at −80 °C in 1.9 mL aliquots in the Augsburg Central Biobank (ACBB) for the duration of the study. Patient material remaining after the completion of the study will be stored at the ACBB for future Ethics Committee-approved research based on the biobank consent.

cfDNA isolation and library preparation for CAPP-seq based analyses

cfDNA extraction and library preparation is performed using the Roche Avenio ctDNA technology (Roche Holding AG, Basel, Switzerland) with a regular input volume of 4 mL plasma. cfDNA concentration is determined using the Qubit 1x dsDNA HS Assay Kit Fluorometer (ThermoFisher Scientific, Waltham, MA, USA). cfDNA quality and size distribution is assessed with the Bioanalyzer High Sensitivity DNA assay (Agilent Technologies, Santa Clara, CA, USA).

The sequencing libraries are prepared according to the manufacturer’s protocol (Roche Avenio ctDNA Surveillance Kit) from the maximum possible input DNA for each sample. DNA concentration of pre-enriched and final libraries are measured with the Qubit 1X dsDNA HS Assay Kit (ThermoFisher Scientific, Waltham, MA, USA). The fragment size and distribution of the libraries are determined using the Bioanalyzer High Sensitivity DNA Analysis (Agilent Technologies, Santa Clara, CA, USA).

Libraries are sequenced using paired-end 100 bp sequencing aiming at a theoretical coverage of 20,000fold on an Illumina Next Seq550 or on an NovaSeq 6000 platform (both Illlumina, San Diego, CA, USA, sequencing performed by the NGS Competence Center Tübingen, NCCT) ensuring a minimum analytical sensitivity of <0.1 %. These analytical sensitivity requirements comply with the Rili-BAEK guidelines, which stipulate a minimum sensitivity of 0.5 % for DNA from cell free body liquids [21].

The Roche Avenio Surveillance Assay is a commercially available assay validated for research use only. In addition, for validation purposes, we performed several preliminary tests with defined mutations (e.g., in the BRAF and EGFR genes) in which different dilutions were systematically compared and checked for validity using digital PCR (QuantStudio™ 3D, ThermoFisher Scientific, Waltham, MA, USA) and the Roche Avenio ctDNA Surveillance Kit.

Collection of clinical data and patient reported outcome measures (PROM), and asservation of germline material, preparation of tissue biopsies, and whole exome sequencing (WES) of TBx are described in Supplementary Methods.

ALPS conceptualization and structure

ALPS was designed as an observational longitudinal study that does not interfere with the patient’s clinical path and monitors the course of therapy at various levels. In addition to obtaining sequential TBx and LBx, these levels also include the documentation of the clinical course with clinical and laboratory parameters, the recording of diagnostic imaging at study inclusion and during sequential staging, and the recording of patient-related outcome measures (PROMs) (Figure 1A). All biosamples and data are collected in a registry platform, which forms the infrastructure level of ALPS, where different study endpoints are addressed and different projects and exploratory objectives defined in the study protocol are set up (Figure 1B).

Figure 1:

ALPS conceptualization, structure and data collection scheme. (A) Patients diagnosed with metastatic or locally advanced incurable tumor disease will be enrolled in the study at the time of diagnosis or change of treatment line and will be followed comprehensively throughout the course of the disease. As outlined in the study protocol, clinical data, radiologic imaging, liquid biopsy (LBx) and tissue biopsy (TBx) samples, and patient-reported outcome measures (PROMs) are systematically collected over time. The follow-up intervals (rpt) are determined by the physicians responsible for the treatment and are approx. 10–12 weeks for most patients. (B) Schematic representation of the structure of the ALPS trial. ALPS is based on an infrastructural level consisting of two pillars (liquid plus tissue biobanking and clinical registry) and on the project level, which is characterized by the primary endpoint, secondary study endpoints as well as a variety of exploratory analyses and respective satellite projects.

In addition to the primary and secondary endpoints defined in the study protocol (concordance/discordance between TBx and LBx, resolution of spatial tumor heterogeneity and monitoring of patients treated with precision medicine), these projects include a number of exploratory projects and endpoints enabled by the open and flexible structure of ALPS. In addition to cfDNA-based genomic analyses, ALPS enables the analysis of other tumor components or information from LBx, such as genome-wide DNA methylation patterns from cfDNA and the isolation and analysis of extracellular vesicles and various peptides including immunopeptidomics (data not shown). Thus, biobanking in ALPS in combination with deep phenotyped data enables future translational research projects and further LBx analyses.

Patient enrollment and clinical characteristics

Recruitment for the ALPS cohort started on 04/01/2021 at the time of initial diagnosis or at the time of change to next treatment line for patients with metastatic or locally advanced, incurable malignant disease. Recruitment is currently open at Augsburg University Hospital for both inpatients and outpatients. Eligible patients are identified by screening appointment lists, clinical records, tumor board recommendations, and direct contact with treating physicians. Biomaterial for TBx is collected during routine clinical biopsy. LBx is collected as part of a routine blood draw. Clinical and patient-centered data are documented and LBx samples are collected synchronously with routine clinical presentations and staging examinations.

The current interim analysis with database cut-off on 10/18/2023 covers a recruitment period of approximately 30 months. During this period, 1,104 patients were screened and ultimately 419 patients who fulfilled the inclusion and exclusion criteria were consented and included in the study (Figure 2). This corresponds to a number needed to screen of 2.5. The screening rate was 28 patients per month from 04/2021 to 04/2022. Through optimizations in the screening procedure, the screening rate was increased to 39 patients per month from 05/2022 until the database cut-off 10/2023. The inclusion rate remained constant over the entire period at approximately 13.8 patients per month (Supplementary Figure 1). Of those screened, 668 patients could not be included in the study. Major reasons for screening errors were failure to meet the in- and exclusion criteria (n=416, 62.3 %), patient refusal (n=106, 15.9 %), loss of contact or loss to follow-up after screening and before inclusion (n=67, 10.0 %), organizational reasons related to the inpatient care of the patient (n=53, 7.9 %), and early death of the patient (n=26, 3.9 %). At time of interim analysis, 179 patients (42.7 %) were active patients in the study protocol. Of the 240 inactive patients, 179 (74.6 %) died and 57 (23.7 %) were lost to follow-up, two declined participation after inclusion and two further were removed a posteriori since criteria for inclusion/exclusion were wrongly assessed.

Figure 2:

ALPS consort diagram showing screening, inclusion, liquid and tissue biopsy rates and patients currently under follow-up as of cut-off Oct 18, 2023.

Patient characteristics are given in Table 1. Thirty-six percent of the included patients are female, median age is 67 years (range 25–91 years). The most common diagnosis in the ALPS cohort is lung cancer (35.1 %) followed by bowel cancer (9.1 %), prostate cancer (7.4 %), and cancer of unknown primary/CUP syndrome (6.2 %, Figure 3A, B). A total of 48 patients from the ALPS cohort (11.5 %) presented to the molecular tumor board (MTB) during their time on study and were thus subject to molecular follow-up via ALPS.

Figure 3:

Age and entity distribution in the ALPS collective. (A) Age distribution in age groups of 10 years for female and male study participants and (B) distribution of cancer entities (according to the MSKCC OncoTree nomenclature) by cut-off Oct 18, 2023.

Longitudinal biobanking and follow-up

Biospecimens are collected at predefined time points based on sequential stagings of each patient. At patient inclusion, timely tissue samples were obtained from 380 patients (90.7 %) and LBx samples from all 419 patients (100 %). Similarly, germline material by means of peripheral blood mononuclear cells (PBMCs) was obtained from the same 419 patients. To reflect the longitudinal course of the patients, LBx samples and clinical data were collected synchronously with the re-staging examinations. For the first 419 patients (comprising a total of 1293 LBx samples), a total of 874 LBx follow-up samples in addition to the baseline LBx samples, are currently available. This corresponds to a median (mean) of 2 (3) (range 1–12) follow-up assessments (Figure 4). At follow-up visit 1, LBx samples were available from 65.2 % of patients, and at follow-up visits 2 and 3, 45.6 and 32.5 %, respectively. For 39.6 % of patients, at least one follow-up TBx (exceeding the re-biopsy rate of 15 % anticipated in the trial protocol) was obtained upon detection of tumor progression. The re-biopsy rate reflects routine procedures in a routine clinical setting. The main reasons for not performing re-biopsies were the risk of the procedure, the deterioration of the patient’s condition and the difficulty of accessing the tumor lesion. The time between follow-up visits differed from patient to patient depending on their clinical follow-up and was 12 weeks in median (range 0–80, interquartile range (IQR): 9–16) at the time of interim analysis. These data represent a “snapshot” as the cohort continues to mature through the 181 patients in follow-up. Concurrent with the sequential preservation of LBx and TBx, we performed the longitudinal clinical documentation, storage of radiological image data, and PROM data.

Figure 4:

Swimmer’s plot depicting the number of total visits (baseline plus follow-up visits) for all individual patients and by cut-off Oct 18, 2023. The Swimmer’s plot shows the number of follow-up examinations for each individual patient by means of horizontal lines separated by vertical blue lines.

Median follow-up visits of the 48 patients who were also discussed in the molecular tumor board (MTB) at Augsburg University Hospital were 4 (range 1–12; IQR 2–6). Here, ALPS offers a unique opportunity to generate not only clinical and patient-centered outcome data but also molecular follow-up for MTB patients who were treated with molecular-guided treatment regimens.

Liquid biobanking, preanalytical quality and cfDNA preparation

The collection and storage of high-quality LBx samples is a central component of ALPS. In order to ensure the pre-analytical quality of the samples, tests with different blood container systems and latencies from sample collection to processing were carried out as part of the study preparation. The decay of hematopoietic cells from peripheral blood between sample collection and processing is a source of “contamination” of cfDNA with genomic DNA (gDNA). Since we minimized the latency time from collection to sample processing to <2 h in the current monocentric setting of ALPS, a cost-effective blood collection could be performed using conventional EDTA containers. A direct comparison with container systems containing cell-stabilizing agents showed similarly low levels of contaminating gDNA for all blood container systems in the period from 1 to about 8 h after collection (Supplementary Figure 2).

For the first 419 patients (comprising a total of 1293 LBx samples), the median time from blood collection to start of sample preparation was 30 min (IQR 21–47 min). Plasma was prepared from five 9 mL EDTA containers at baseline and from three 9 mL EDTA containers for follow-up visits. The median time from blood collection to cryoconservation of plasma samples was 90 min (IQR 72–111 min). Plasma samples were stored in aliquots of 1.9 mL at −80 °C.

So far, cfDNA was isolated from 330 LBx samples. The concentration of cfDNA varied significantly between entities, individual patients, samples, and time points. On average, 12.0 ng/mL plasma (range 1.2–2300.0 ng/mL plasma) could be isolated. Figure 5A shows the concentration at the time of study inclusion (baseline) differentiated by individual entities. During follow-up, cfDNA was also successfully detected in subsequent LBx samples, although the amount of cfDNA isolated here varied greatly between individuals and timepoints (Figure 5B).

Figure 5:

Overview of cfDNA quantity in ALPS. (A) Concentration of cell-free DNA (cfDNA) in ng cfDNA per 1 mL plasma from plasma isolations of the first 126 patients, grouped by cancer entity. (B) Changes in the cfDNA quantities isolated from individual patients over the course of the follow-up visits. The distribution of cfDNA concentrations of the first 126 patients was divided into quartiles at the time of study inclusion (with Q1of 6.9 ng and Q3 of 45.4 ng cfDNA/ml plasma). These quartile-related cut-off values were used for the distribution of the isolated cfDNA quantities at the follow-up visits in order to visualize cfDNA dynamics and relevant differences in cfDNA recovery.

CAPP-seq analysis was successfully performed on the cfDNA of the first 48 patients and demonstrated that the isolated material was suitable for the planned analyses and endpoints and could be processed accordingly. The sequencing parameters were chosen with 60x 10⁶ reads per sample based on an on-target rate of 40–60 % and a duplicate rate of 20–30 %. The aim was to generate an average unique coverage of 6,000–7,000-fold to achieve a sensitivity of approximately 0.1 % for the detection of variants. CAPP-seq analyses was performed for the first 110 samples from 60 patients and yielded between 3.93 and 29.72 million reads with an on-target rate of 40–80 %. After first level processing and deduplication of the sequencing data, we achieved an average read depth of 5,218 reads (range 1,390–9,132) and a detection sensitivity for variants of <0.1 % per sample, demonstrating the applicability of CAPP-seq to samples of the ALPS cohort and the feasibility of our approach. In four cases, the median sequencing depth was below 3,000 (2,475). In these patients, it could not be guaranteed that the sample sensitivity reached the target detection sensitivity of 0.1 % as described above (Supplementary Figure 3A, B).

Using this approach, different numbers of cancer-related mutations and copy number variations were identified in all analyzed samples from five subcohorts defined by common entities (Figure 6A). These could be used to track tumor mass by estimating allelic burden and indicate clonal evolution under therapy. Figure 6B shows an example of the integration of LBx-derived data on the variant allele frequency of individual variants and the resulting tumor burden, radiological imaging data and the course of treatment.

Figure 6:

CAPP-seq based ctDNA analyses in ALPS. (A) Filtered variants by major entities. The entire spectrum of all entities is discussed in more detail in Figure 3B. In this illustration, CUP patients are listed in particular under “other”. (B) Example of a patient with non-small cell lung cancer (NSCLC) who was monitored in ALPS from the start of first-line therapy. The upper panel shows CT-based imaging before initiation of therapy (t0), at time t1 after 14 weeks (wks) of immunochemotherapy and at time t2 after a further 14 weeks of immune maintenance therapy. Response (PR, partial response) is indicated according to RECIST criteria (v1.1). The white arrows mark the target lesion. The mean panel shows the variant allele frequency (VAF) in percent over timepoints t0-t2 in the genes/gene regions with detectable variants. Maximum detection sensitivity for variants is 0.1 %. A VAF of 0 is considered to be equivalent to not detected at a sensitivity level of 0.1 %. The lower panel depicts the course of therapy with the following substances: Carbo (carboplatin), Pac (paclitaxel), Pembro (pembrolizumab).